A Single Multiplex crRNA Array for FnCpf1-Mediated Human Genome Editing

Abstract

Cpf1 has been harnessed as a tool for genome manipulation in various species because of its simplicity and high efficiency. Our recent study demonstrated that FnCpf1 could be utilized for human genome editing with notable advantages for target sequence selection due to the flexibility of the protospacer adjacent motif (PAM) sequence. Multiplex genome editing provides a powerful tool for targeting members of multigene families, dissecting gene networks, modeling multigenic disorders in vivo, and applying gene therapy. However, there are no reports at present that show FnCpf1-mediated multiplex genome editing via a single customized CRISPR RNA (crRNA) array. In the present study, we utilize a single customized crRNA array to simultaneously target multiple genes in human cells. In addition, we also demonstrate that a single customized crRNA array to target multiple sites in one gene could be achieved. Collectively, FnCpf1, a powerful genome-editing tool for multiple genomic targets, can be harnessed for effective manipulation of the human genome.

The authors utilized a single customized crRNA array of FnCpf1-mediated gene editing to simultaneously target multiple genes or multiple sites in one gene in human cells. Their results demonstrated FnCpf1 can be harnessed for effective manipulation of multiple human genomic targets.

Introduction

Cpf1 is a characterized effector endonuclease protein that is a class II member of the CRISPR-Cas family. Previous study suggests it is a highly specific programmable nuclease.¹ Compared with Cas9, the most commonly used genome-editing enzyme, Cpf1 has a notable feature that it requires only a single CRISPR RNA (crRNA) without the assistance of trans-activating crRNA (tracrRNA). In bacteria, Cas9 is dependent on RNase III to excise crRNAs from a CRISPR array. In contrast, Cpf1 is a dual nuclease that not only cleaves target genes complementary to the crRNA guide but also processes the crRNA array into mature crRNAs using its own RNase activity.¹ Moreover, unlike Cas9, which creates blunt-ended cleavage products proximal to the protospacer adjacent motif (PAM) site, Cpf1 generates a staggered double-strand break (DSB) resulting in 5′ overhangs distal to the PAM site. Cpf1 prefers a thymine-rich PAM sequence near the target sites, which increases the availability of CRISPR-endonuclease-editable genomic sites, yet the PAM for Streptococcus pyogenes Cas9 (SpCas9) and Staphylococcus aureus Cas9 (SaCas9) are 5′-NGG-3′ and 5′-NNGRRT-3′, respectively. Due to its unique PAM selection characteristics, Cpf1 has emerged as an additional powerful tool for genome editing that can facilitate genetic functional studies, and it has potential therapeutic applications. Until now, AsCpf1 (Acidaminococcus sp. BV3L6 Cpf1), LbCpf1 (Lachnospiraceae bacterium ND2006 Cpf1), and FnCpf1 (Francisellanovicida U112. Cpf1) have been applied successfully to gene modification in various species, including human, mouse, zebrafish, Xenopus, and plants; also, they have been investigated for preclinical studies with models of human disease.2, 3, 4, 5

Multiplex genome editing provides a powerful tool for targeting members of multigene families, dissecting gene networks, modeling multigenic disorders in vivo, and applying gene therapy. With CRISPR/Cas9, multiplex genome editing could be achieved. In principle, the most common strategy is to stack multiple single-guide RNA (sgRNA)-expressing cassettes in one plasmid construct or to use multiple constructs for multiplex genome editing.6, 7, 8 In addition, an advanced strategy would be to compact a cluster of gRNAs with different spacers in one synthetic gene and engineer an RNA-processing system to excise individual gRNAs from the transcript.9, 10 Although multiplex gene editing with CRISPR/Cas9 has been reported, it requires large constructs to express multiple sgRNA cassettes, is labor intensive, and can be costly.11, 12 Recently, the natural structure of the Cpf1 system has been adopted for multiplex gene editing or gene regulation in mammalian cells.13, 14 Simultaneously editing up to four genes has been achieved via a single crRNA array spaced by mature direct repeats (DRs), with AsCpf1 and LbCpf1.¹⁴ However, it remains a challenge to apply this approach if the target loci have no suitable PAM sequences, which are limiting for these enzymes.

Recently, we reported that FnCpf1 possesses activity in human cells and recognizes a more compatible PAM (5′-KYTV-3′)⁴ compared with AsCpf1 and LbCpf1, which recognize the canonical 5′-TTTN-3′ PAM.4, 15 Here we sought to investigate and optimize a system for FnCpf1-mediated gene targeting in human cells via a single multiplex crRNA array.

Results

Advantages of a Single crRNA Array for FnCpf1-Mediated Multiplex Gene Editing

Previous studies showed that AsCpf1, LbCpf1, and FnCpf1 could trigger genome editing in human cells.1, 4 Specifically, with a 21-nt guide segment, FnCpf1 possesses higher activity than with a 23-nt guide segment.⁴ By contrast, it has been reported that AsCpf1 possesses a robust activity with a 23-nt guide segment.¹ Furthermore, AsCpf1 and LbCpf1 have been adopted for multiple gene editing in mammalian cells, where up to four genes can be simultaneously edited with a single crRNA array spaced by 19-nt mature DRs. To perform comparative studies, we designed 21-nt and 23-nt guide sequences for FnCpf1 and AsCpf1, and we quantified GFP-negative cells (GFP knockout) using flow cytometry to track their performance. First, we directly synthesized the corresponding DNA fragments for the single crRNA array to target five genes (GFP, EMX1, HBB, CCR5, and VEGFA) with 21- or 23-nt spacer sequences (Table S1). To test the choice of PAM sequence, we generated plasmids to target three sites with the different PAM sequences in GFP (5′-TTTA-3′, 5′-GTTC-3′, and 5′-GCTA-3′; Figure 1A). We transfected the plasmids into 293-SC1, a cell line expressing GFP generated by lentiviral transduction, and we monitored the efficiency of GFP editing (Figure 1B). Unsurprisingly, both FnCpf1 and AsCpf1 with canonical 5′-TTTA-3′ PAM sequences successfully triggered GFP gene inactivation, typically with a 21-nt spacer sequence; the efficiencies were 37% and 24%, respectively. With 5′-GTTC-3′ PAM, the mutation frequencies triggered by FnCpf1 and AsCpf1 were 36.3% and 15.2%, respectively (Figure 1C).

Figure 1.

FnCpf1 and AsCpf1 Mediated Multiplex Gene Editing in Human Cells

(A) Schematic of multiplex gene editing with Cpf1. Genome editing at five different genomic loci mediated by FnCpf1 and AsCpf1 with different arrays (array 1, crRNAs in their mature form [19-nt DR with 21-nt guides]; array 2, crRNAs in their mature form [19-nt DR with 23-nt guides]). Schematic diagram of targeted sites with different PAMs in GFP. PAM sequences and 21-nt target sites are shown in pink and black, respectively. 23-nt target sites contain the additional two bases shown in blue. (B) Schematic of the GFP-reporter system for measuring Cpf1-mediated DNA cleavage in human cells. (C) The activity of FnCpf1 and AsCpf1 with different lengths of guide sequences within the three targeted GFP sites. (D) Gene-editing efficiency at HBB, CCR5, EMX1, and VEGFA loci with FnCpf1 and AsCpf1 was measured by the T7E1 assay. (E) The activity of Cpf1 (FnCpf1 and AsCpf1) with exchanged DR at GFP sites 1 and 2 by flow cytometry. hU6, human U6; CMV, cytomegalovirus promoter; DR, direct repeat sequence; NC, negative control. Error bars, SEM; n = 3; *p < 0.05 and **p < 0.01.

As we expected, AsCpf1 containing 5′-GCTA-3′ PAM sequences failed to cleave GFP sites, both with 21-nt and 23-nt spacers (the average efficiencies were 6.5% and 5.3%, respectively), while FnCpf1 triggered robust cleavage activity (the efficiencies were 23.76% and 23.31%, respectively). Taken together, these data showed that a single crRNA targeting GFP or targeting multiple genes shared the same pattern, and there was a notable target selection advantage of FnCpf1 due to its PAM specificity (Figure 1C).

Then we sought to investigate the targeting efficiency of additional target genes, including four endogenous genes (HBB, CCR5, EMX1, and VEGFA). The PAM sequences for three genes (EMX1, HBB, and CCR5) are 5′-TTTN-3′, which are compatible for both AsCpf1 and FnCpf1. For VEGFA, the PAM sequence is 5′-GTTT-3′ and it is only compatible with FnCpf1, which may distinguish the effects of the PAM between AsCpf1 and FnCpf1. Notably, FnCpf1 possessed robust activity toward HBB, CCR5, and EMX1, with cleavage activity of up to 30.7%, 24.7%, and 31.5%. As for AsCpf1, the efficiencies were 15.3%, 30.37%, and 20.8%, respectively (Figure 1D). For the VEGFA gene, we could clearly detect the cleavage activity of FnCpf1, but no activity of AsCpf1 was observed. These data demonstrated that PAM is advantageous in single crRNA array involving FnCpf1-mediated multiple gene editing.

Promoters and DR Sequence Effects for Single crRNA Array

In most cases, Cpf1 relies on RNase III to excise crRNAs from a CRISPR array.¹⁶ To further determine whether crRNA excised from different promoter-driven transcripts could be exploited to enhance gene-editing efficiency in our system, we generated plasmids harboring different promoters. This enabled us to assess the efficiency of genome editing with crRNA driven from human-U6 (hU6), mouse-U6 (mU6), cytomegalovirus (CMV)-enhancer-human-U6 (ChU6), and CMV promoters with different poly(A) sequences (Table S2). To measure the KO (knockout) efficiency of GFP, we quantified GFP-negative cells using flow cytometry. These results demonstrated that the ability of FnCpf1 to excise crRNA from transcripts by mU6 (35.3%) and ChU6 (32.6%) had the same or slightly higher efficiency compared with parental plasmids harboring hU6 (33.9%). In contrast, the efficiencies of editing mediated by CMV promoter-expressed crRNAs with different poly(A) sequences were 8.9%, 10.7%, and 10.4%, respectively (Figure S1). Here we showed that, although mU6 and ChU6 activities can’t be increased with the replacement of the promoter for crRNA, alternative promoters may be utilized for specific applications.

Our previous results showed that DRs of Cpf1 are interchangeable.⁴ In the case of multiple gene editing via a single crRNA array, we speculated that the same pattern may be observed with FnCpf1 (Figure S2A). To investigate the specificity of DR sequences, we interchanged the crRNA cassette with a 21-nt and 23-nt spacer sequence between FnCpf1 and AsCpf1. The data demonstrated that FnCpf1-AsDR and AsCpf1-FnDR systems, compared with their canonical ones, caused decreased or similar efficiency at the GFP site 1 (19.2% versus 28.8% and 12.9% versus 22.0%) and site 2 (24.7% versus 38.7% and 15.1% versus 15.0%; Figures 1E and S2B). In addition, the effect of both interchanged Cpf1 systems with a 23-nt spacer sequence was also consistent with the above results (Figures 1E and S2C). It revealed that non-canonical DR weakened RNA cleavage. Taken together, these results revealed that, in the case of multiplex gene editing using single crRNA arrays, the promoter for driving crRNA may be changeable, highlighting the broad flexibility of FnCpf1. What’s more, we observed that the DR is relatively highly conserved and specific for Cpf1 systems.

Efficiency of Introducing Target Gene Mutations by FnCpf1

Next, we compared the activity of the single crRNA array in targeting a single gene and multiple genes. We selected three GFP sites for the comparison, using plasmid amounts of 250, 500, 750, and 1,000 ng in each transfection (Table S3). The editing efficiency of the multiplex array mediated appeared to be higher than the single site at all doses of three sites, particularly with the 500-ng plasmid (Figure 2A). To confirm it, four endogenous genes were investigated by T7 Endonuclease 1 assay analysis. The efficiencies of FnCpf1 with the multiplex array showed similar or higher modification compared with a single array (Figure S3). Collectively, this showed here that a single crRNA array targeting multiple genes or a single gene site may have similar activity.

Figure 2.

FnCpf1 Mediated Single and Multiple Gene Editing in Human Cells

(A) Comparison of the activity between single and multiple gene editing using different doses of crRNA expression plasmids (250, 500, 750, and 1,000 ng) targeting three sites of one gene GFP. (B) Schematic of the analysis strategy of indel events in clonally selected colonies by Sanger sequencing of the fragments harboring the targeted sites. (C) Quantification of indel events identified with Sanger sequencing in clonally selected colonies. The upper panel represents single gene editing and the lower panel represents multiple gene editing. Each column represents one clonally selected colony; purple rectangles indicate target genes with all alleles edited. (D) Analysis of the ratio of multiple gene targeting efficiency in individual cells. Error bars, SEM; n = 3; **p < 0.01.

It has been reported that most of the non-homologous end joining (NHEJ) mutations derived from Cpf1 are predominantly short insertions or deletions (indels), most commonly 3–30 bp in size. While, with the single crRNA array targeting multiple genes, we want to know if it would still be the same. To address this, the mutation pattern triggered by FnCpf1 with a crRNA array was mapped (Figures 2B and 2C). Overall, the mutation efficiencies at different sites in EMX1, HBB, and CCR5 were relatively high and slightly lower at the VEGFA site (Figure 2C). Specifically, most of the mutations in targeted regions were heterozygous indels, with three homozygous deletions (two for EMX1 and one for HBB; Figure S4). Notably, triple mutations and quadruple mutations accounted for 10% and 20% of the sequences analyzed, respectively (Figure 2D).

Next we sought to determine the off-target effects of FnCpf1-mediated multiplex gene targeting via a single crRNA array. Online software was used for the prediction of off-target effects (http://www.rgenome.net/cas-offinder/). We assayed potential off-target sites for each target by T7E1. We found slightly off-target effects compared with on-target effects (Table S4; Figure S5). Precise genome analysis tools, i.e., whole-genome sequencing (WGS), are needed to characterize additional potential off-target effects. Collectively, these results demonstrated that, with the utility of a single multiplex array, FnCpf1 could be harnessed as an efficient tool for multiple gene KO.

Using the FnCpf1 System to Target Multiple Sites in One Gene

Given that the results described above showed that the efficiency of an array targeting multiple sites in one gene appears to offer higher activity over the single crRNA targeting only one site, we speculated that this strategy can be used for functional gene exploration. To further assess this, we designed a tandem array harboring four spacer sequences to target two genes, one for GFP for tracking genome-editing efficiency and the other three sequences targeting MAPK14 encoding the components of the P38 pathway. To confirm the results, another gene, NOTCH1, was selected (Figure 3A; Table S5). As an initial test, the efficiency in 293-SC1 cells was analyzed by flow cytometry. The results showed the efficiency of GFP disruption of each construct was similar (Figure S6A). To increase the KO efficiency, a puromycin resistance gene was added to the vector, which conferred a selective advantage on puromycin-treated cells that was dramatically enriched (Figures 3A and S6B). To assess KO efficiency we utilized western blotting. As expected, the results showed that each protein’s expression levels of MAPK14 and NOTCH1 were decreased significantly (Figure 3B). By comparison, the protein level of MAPK14 with multiple crRNA (S1–S3, 24.5%) was dramatically lower than the single crRNA (S1, 55.2%; S2, 68.8%; S3, 55.8%). As for NOTCH1, the rates were 50.4% (S1–S3), 78.8% (S1), 72.3% (S2), and 71.7% (S3), respectively. The cleavage of multiple gene editing was further confirmed by PCR products cloning and sequencing. The majority of the mutations observed were small deletions (Figures 3C and S7A). In addition, the differential activities of various spacer sequences were observed (Figures S7B and S7C). Collectively, here we showed that the multiplex FnCpf1 system targeting multiple sites in one gene may have advantages over a single targeting site for functional studies.

Figure 3.

The Efficiency of FnCpf1 System Targeting of Multiple Sites in One Gene

(A) The structure illustrates the multiplex expression array of GFP site 1 and three additional targeted sites, each targeted to the same endogenous genes containing MAPK14 and NOTCH1. (B) Protein levels of MAPK14 or NOTCH1 genes via treatment with single and multiple gene editing. (C) Representative mutation patterns of MAPK14 and NOTCH1. PCR products were cloned into pJET1.2 vector and sequenced with Sanger sequencing. Dashes represent DNA deletions and blue is used to represent DNA insertions. DR, direct repeat sequence; S1, spacer sequence 1; S2, spacer sequence 2; S3, spacer sequence 3; NC, negative control.

Discussion

Cpf1-mediated genome editing has been previously performed in mammalian cells with high fidelity.1, 17, 18 In the present study, we show a proof of principle of multiplex FnCpf1-mediated gene KO in multiple genes or multiple sites in a single gene using a single crRNA array in human cells. According to previous studies, 80% of human exons are <200 bp in length;¹⁹ therefore, very few target sequences are available for selection with AsCpf1 and LbCpf1 due to the restricted PAM compatibility with these enzymes.¹ Fortunately, this limitation can largely be resolved with the discovery that FnCpf1 has a much less restrictive PAM specificity and possesses efficient cleavage activity in human cells.⁴ For example, in GFP, there are only three sites harboring a 5′-TTTN-3′ PAM sequence that could be used for AsCpf1 and LbCpf1 cleavage, including one near the stop codon, while 50 5′-KYTV-3′ PAM sites are compatible with FnCpf1.

In addition, we found that additional promoters, including mU6 and ChU6, also could be utilized to drive crRNA expression and achieve desirable cleavage activity. However, it’s still unknown why the RNA polymerase II promoter (CMV) has no detectable activity in our study, which is not consistent with the literature.²⁰ Further studies need to be performed to elucidate the underlying reasons for this discrepancy.

With this system, 6% gene KO (three to 50 colonies) with no selection could be obtained, and these were homozygous at some sites (two for EMX1 and one for HBB). Notably, triple mutations and quadruple mutations were identified in 10% and 20% of cases, respectively. The results are of importance for the application of FnCpf1 gene editing in drug screening, cellular or animal disease models, and the dissection of gene function.10, 17 For example, after the generation of a double-allele KO of transformation-related protein 53 (Trp53), the specific cell line could be used for screening small molecules for restoring Trp53-KO-induced genome instability.²¹

It is highly encouraging that the expression of some genes is dramatically decreased when we designed and tested multiple targeting sites simultaneously in a single gene (Figure 3). Based on our previous study, we found significant differences in activity between different genes and even with the same gene with different targeting sites.⁴ Using this approach, we have obtained a high efficiency of gene KO when using three to five sites to target one gene. Also, this approach may not depend on testing the cleavage activity of each target sequence. At present, we do not know the double-allele KO efficiency of individual colonies after selection. In addition, we speculate that its efficiency should be relatively high because protein expression level decreased significantly via targeting multiple sites in a single gene. In this study, the complete indel information has not yet been assessed fully, although it has been investigated via cloning and sequencing the fragments harboring the targeting site. Not surprisingly, we observed that each site is targeted with a different efficiency, consistent with previous results of single crRNA targeting.¹⁴ As for the total indel number, the NHEJ triggered by the combination of targeting sites in the array may lead to large deletions in proximity to these target sites, which may be underestimated by PCR-based sequencing.

FnCpf1-mediated multiplex genome editing also has additional potential uses for the manipulation of the genome. For example, inactivated FnCpf1 fused with VPR activator (consisting of the herpes simplex virus-derived VP16 activator, the human nuclear factor κB (NF-κB)-p65 activation domain, and the Epstein-Barr virus-derived R transactivator) may provide another tool for enhancing multiplex target gene expression.²² Also, an inactivated mutant FnCpf1 plus a single base-editing enzyme or DNA methylation or demethylation enzyme may be used for inserting both multiplex point mutations, as well as even correcting regulation of targeted gene expression.23, 24, 25 Taken together, we demonstrated that FnCpf1-mediated multiplex genome editing is a powerful human genome modification tool.

Materials and Methods

Plasmids Encoding Cpf1 and crRNA

Plasmids for the expression of Francisellanovicida U112 (FnCpf1, PY004) and Acidaminococcus sp. BV3L6 (AsCpf1, PY094) were obtained from Addgene (Addgene plasmids 69976 and 84743, respectively). To construct the FnCpf1 multiple array, the multiplex crRNA expression cassettes were synthesized or amplified by overlap PCR of PCR-based accurate synthesis (PAS) protocol for each crRNA;²⁶ the utilized oligonucleotides are summarized in Table S6. To synthesize DNA sequences, oligonucleotides were designed using online software (http://54.235.254.95/gd/). Oligonucleotides were mixed at an equal ratio and amplified by super-fidelity DNA polymerase (Vazyme Biotech, Nanjing, China). The reaction mix was set up as a protocol, and PCR was performed using the following cycling conditions: 94°C for 30 s, 60°C for 45 s, and 72°C for 50 s × 32 cycles and 72°C for 10 min (Bio-Rad, CA). Purified PCR products and vectors were treated with the BsmbI (New England Biolabs) for the ligation. Plasmid DNA and genomic DNA were isolated by standard techniques. DNA sequencing confirmed the desired sequence in the constructs.

Cells and Cell Culture

HEK293 cells were obtained from ATCC (CRL-1573), and they were grown at 37°C in 5% CO₂ in DMEM (Gibco by Life Technologies, Grand Island, NY) and 10% heat-inactivated fetal bovine serum (Gibco by Life Technologies, Grand Island, NY). HEK293 cells expressing GFP were generated by lentiviral transduction as previously described.²⁷ Drug-resistant individual colonies of transduced HEK293 cells were isolated and one of them named 293-SC1. To maintain GFP expression, the 293-SC1 culture medium contained puromycin.

On day 0, 3.4 × 10⁵ HEK293 cells were seeded in six-well plates. One day 1, the cells were transfected with plasmid using TurboFect Transfection Reagent (Thermo Fisher Scientific, MA). The cells were harvested at day 10∼14. Indel information was identified via CloneJET PCR Cloning Kit (Thermo Fisher Scientific, MA) and Sanger sequencing. The protein levels of the gene expression were determined by western blotting.

Flow Cytometry Analysis

The flow cytometry protocol was described previously.²⁷ Briefly, on day 0, 1.7 × 10⁵ 293-SC1 cells/well were seeded in 12-well plates. On day 1, the cells were transfected with plasmids using Transfection Reagent (Thermo Fisher Scientific, MA). Fresh medium was added on day 2. On day 3, cells were washed with PBS, trypsinized, and harvested for genomic DNAs isolation or flow cytometry. Specifically, GFP KO efficiency was assessed via flow cytometry (BD Biosciences, NY) and CellQuest software was used to analyze data.

T7E1 Nuclease Assay for Genome Editing

Fragments harboring the indel mutations were amplified by PCR using the primer sets listed in Table S3. For T7E1 (New England Biolabs) assay, 300 ng purified PCR products were mixed with 1.5 μL 10 × NEB#2 buffer and ultrapure water to a final volume of 14.5 μL, and they were subjected to re-annealing to enable heteroduplex formation. After re-annealing, products were treated with T7E1 0.5 μL. Indel percentage was determined by the formula 100 × (1 − sqrt (1 − (b + c) / (a + b + c))), which was described previously.⁴

Off-Target Analysis for FnCpf1

We examined the possibility that FnCpf1 induced off-target mutations. The potential off-target sites were predicted using online software (http://www.rgenome.net/cas-offinder/). The fragments harboring potential off-target sites were amplified (primer information in Table S3) and digested by T7E1.

Western Blotting Analysis

To assess genome-editing efficiency, HEK293 cell were transfected with plasmids of targeting genes containing MAPK14 and NOTCH1 for P38 and NOTCH1 signal pathways, respectively. We added puromycin at day 2. Cells were harvested at day 10∼14. Western blotting was performed by standard protocol. Primary antibodies were used from Cell Signaling Technology (GAPDH [D16H11] XP Rabbit mAb, P38 MAPK [D13E1] XP Rabbit mAb, and Notch1 [D1E11] XP Rabbit mAb; Cell Signaling Technology, Boston, MA). Goat anti-rabbit IgG (H+L) (LI-COR Biosciences, Lincoln, NE) was used as the secondary antibody.

Statistics

Statistical analysis was conducted with SPSS18.0 software. All data are expressed as mean ± SEM. Differences were determined by two-tailed Student’s t test between two groups. The criterion for statistical significance was *p < 0.05 or **p < 0.01.

Author Contributions

F.G. conceived the idea. H.S., F.L., J.L., F.Y., Z. Zeng, X.L., and M.T. performed the experiments. X.G., C.L., J.Z., Z. Zhang, J.Q., Z.S., and F.G. performed data analyses and F.G. wrote the manuscript. All authors have read and approved the final manuscript.

Conflicts of Interest

The authors declare no competing interests.